A Generative AI Engineer has built an LLM application using a pay-per-token Foundation Model API. As they prepare for production deployment, how can they ensure the model endpoint can handle high volumes of incoming requests?